Beruflich Dokumente
Kultur Dokumente
www.elsevier.com/locate/ndteint
Abstract
Automatic inspection of welded gas pipelines is desirable because human inspectors are not always consistent evaluators. In addition,
automatic inspection decreases the cost of inspection process and improves the inspection quality. In this paper, an expert vision system for
automatic inspection of gas pipeline welding defects from radiographic films is presented. The proposed system has been established in the
Metrology lab, Mansoura University, Faculty of Engineering. The software, named AutoWDI, is fully written in lab using Microsoft Visual
Cþ þ and is ready to run on any Windows environment. The proposed vision system is used to capture images for the radiographic films then
applies various image processing and computer vision algorithms to recognize the defects and to make acceptance decisions according to
international standards. The expert system is based on a knowledge base, which was gathered from specialists, textbooks and international
standards. The proposed system is capable of identifying and testing the main types of welding defects (11 defects) in gas pipelines welded by
shielded metal arc welding.
q 2003 Elsevier Ltd. All rights reserved.
Keywords: Expert system; Welding defects; Quality control; Computer vision; Radiography
Weld (CW), Edge of Weld (EW) and Base Metal (BM). 5.3. The acceptance decision algorithm
For defects, which lie at the EW and the CW regions, a
tolerance zone equal to 5% of the weld height (size in Y The Acceptance Decision Algorithm (ADA) is used to
direction) is taken above and below both the edge of make an acceptance decision about the identified defects.
weld and the center of weld, respectively. The width of After the defect type is identified, its dimensions are
weld is calculated from the window specified by the user. compared with international standards to make the accep-
The location of each defect could be estimated by tance decision. The proposed system supports the most
comparing the coordinates of the minimum rectangle international standards codes (API 1104, ASME, DIN, BS,
enclosing the defect with the coordinates of the four ABS, AWS, and JIS) for the acceptance criteria of weld
mentioned regions. defects. In addition, any new code can be added easily to the
system through a user-friendly dialog box. The default code
is the American Petroleum Institute (API).
6. Experimental study
the form factor ðFf Þ; the rectangularity factor ðRf Þ; shape, has been successfully defined and the proposed algorithms
orientation and the location. (DIA and ADA) are capable of identifying and inspecting all
defects, which are predefined by the identification tree.
Using such expert system could decrease the cost of the
7. Discussions inspection process by using human inspectors with reason-
able knowledge instead of specialists for the general
The results obtained by both the proposed system and the inspection processes. In addition, the enhancement algo-
specialists proved that the identification tree of the defects rithms provided with the proposed system could reduce
Fig. 6. Example for detecting and identifying irregular defects (Burn Through).
306 H.I. Shafeek et al. / NDT&E International 37 (2004) 301–307
Fig. 7. Example for detecting and identifying elongated (horizontal) defects (Incomplete Fusion).
Fig. 8. Example for detecting and identifying elongated (vertical) defects (Transverse Crack).
the inspection time, especially for defects, which are not in radiographic films. Two algorithms were introduced for
clear in the radiographic films. the identification and inspection process. The first algorithm
(DIA) is used to identify the type of defects based on a
predefined identification tree, which was gathered from
8. Conclusions specialists, textbooks and international standards. The
second algorithm (ADA) is used to make an acceptance
An expert vision system for the automatic identification decision for the identified defects according to international
and inspection of gas pipeline welding defects has been standards. In addition, any new standards or customized
introduced. The system is capable of identifying and codes could be added to the system. The inspected defects
inspecting the most common welding defects (11 defects) were identified and inspected correctly as reported by
H.I. Shafeek et al. / NDT&E International 37 (2004) 301–307 307
the specialists of weld inspection. The proposed system is Structural Integrity for Nuclear and Pressurized Components; 1998.
considered quite cheap compared with the commercial p. 20 –22.
[5] Litsikas K. Computer power changes the face of NDT. J Quality
automatic inspection systems and eliminates the need for [QUA] 1996;35:36 –8.
image interoperation by a skilled inspectors. [6] Williams H, Barbian OA. Operational experience with inline
ultrasonic crack inspection of German crude oil pipelines. In:
Proceedings of Seventh European conference on Non-Destructive
Testing; 1998, 3(8).
9. Future work [7] Knell M. Improving actual film quality using image quality
parameters from new national and international standards, vol. 3.
Currently, we are working in a project to apply the In: Proceedings of 14th World conference on Non-Destructive
proposed system in the interpretation and inspection of Testing, New Delhi; 1996. p. 1351–6.
[8] Rao BPC. Visual techniques in non-destructive testing, encyclopedia
radiographic films in the field. The aim of this work is to test
of materials: science and technology. Amsterdam: Elsevier; 2001. p.
and develop the proposed system so that it can be used to 6043– 6.
store the inspection results of radiographic films to a [9] Fletcher MJ. Fully automatic inspection of welds now a reality. NDT
database system. In addition, other information about the E Int 1996;29(3):191.
inspected welds and the welders will be stored to the [10] Awcock GW, Thomas R. Applied image processing. New York:
McGraw-Hill; 1996. p. 228– 30.
database for further analysis in the company such as
[11] Ho SK, White RM, Lucas J. A vision system for automated crack
evaluating the welders’ performance and the common detection in welds. NDT E Int 1996;29(2):129.
defects produced by each welder. [12] Montini G, Bordini MG, Boschetti F, Ripamonti S. EXACT: an expert
system for real time X-ray image analysis and classification of
weldings. NDT E Int 1996;29(3):184.
[13] Elewa IM, Abdel-Shafy AA, Gadelmawla ES and Shafeek HI.
References Assessment of welding defects for gas pipeline radiographs using
computer vision. Proceedings of Third Assuit International Con-
[1] Foulds JR. Concerns about seam-welded piping at elevated tempera- ference on Mechanical Engineering Advanced Technology for
tures. Proc SPIE 1996;2947:67–77. Industrial Production, Assiut, Egypt; 2002. p. 216–23.
[2] Rao BPC. Advances in non-destructive inspection and evaluation of [14] Perner P, Zscherpel U, Jacobsen C. A comparison between neural
welds. Proceedings of the Workshop on Welding Technology for networks and decision trees based on data from industrial radio-
Engineers Organized by Indian Institute of Welding, Chennai, India; graphic testing. Pattern Recognition Lett 2002;48(1):47 –54.
2002. p. 71–90. [15] Wang G, Warren LT. Automatic identification of different types of
[3] Chan C, Grantham KH. Fabric defect detection by Fourier analysis. welding defects in radiographic images. NDT E Int 2002;35(8):
IEEE Trans Ind Appl 2000;36(5):1267–75. 519–28.
[4] Yoneyama H, Aono K, Nakada S, Igeta M. A study of capability [16] Lashkia V. Defect detection in X-ray images using fuzzy reasoning.
of flaw detection and accuracy of flaw sizing using ultrasonic Image Vision Comput 2001;19(5):261–9.
testing. First International Conference on NDE in Relation to [17] Parker JR. Practical computer vision using C. Printic Welley; 1995.